In the current 'make install' (but used to be only 'make install-dev') we install a pile of PMC .dump files. Originally, this was because some languages had dynpmcs that subclassed core PMCs, and the tools needed these files for meta-data to build the subclass. Two questions:
- Do the new PMC-building tools still use the installed .dump files of parents for subclassing? Are all languages using the new tools? - If the installed .dump files are still used, can we change the process so we generate a small .json file of the necessary meta-data for each core PMC during the build process, and then only install small data files instead of the full .dump files? The mass of installed mostly useless files has been bugging me for a while, but this solution only just occurred to me. 9 MB of .dump files may not seem like much, until you realize that our total install size is 39 MB. Of the remaining bulk, we've got: - 14 MB for libparrot.a - 8 MB for libparrot.so.<version> - 4 MB for lib/parrot/<version>/library (PIR libraries, and their compiled PBC) - 2 MB for lib/parrot/<version>/tools (Perl libraries, Perl and PIR scripts) Some of this is definitely trimmable. (I'm thinking of embedded environments, distro install CDs, mobile phones, plug computers, etc. where a light footprint is a killer feature.) Allison _______________________________________________ http://lists.parrot.org/mailman/listinfo/parrot-dev
